kstudiochat 1.1.0

K-CHAT — Universal Chatbot Engine. Anti-hallucination by construction.

K-CHAT — Universal Chatbot Engine

Zero-hallucination chatbot engine. Folder-based. No ML expertise needed.

pip install kstudiochat
kstudiochat init ./bot --template restaurant
kstudiochat chat --data ./bot

Three commands. Working bot. No API key required (SimulatedLLM built-in).

---

Quick Start

# Install
pip install kstudiochat

# Create a bot from a pre-built template
kstudiochat init ./bot --template government

# Chat (uses SimulatedLLM — offline, no API key)
kstudiochat chat --data ./bot

# Or use a real LLM
pip install kstudiochat[openai]
# Edit bot/config.json → llm.provider: "openai" | "deepseek" | "mimo"

Templates

Template	Description
government	FAQ for passports, permits, taxes
restaurant	Menu, hours, reservations
healthcare	General health info, clinic FAQ

---

How It Works

Drop a folder. Get a chatbot. No code.

my-bot/
├── config.json       # Bot identity, LLM provider
├── knowledge/        # Documents the bot answers from
│   ├── faq.md
│   └── pricing.md
├── SOUL.md           # Personality (optional)
├── intents.json      # Keywords (optional)
└── refusal.json      # Fallback messages (optional)

Pipeline

User → Sanitize → Ethics → Intent → Feed ALL data → LLM → Verify → Response

Three-gate prompt:

• Gate 1 — Greeting/thanks/farewell → LLM without data (natural)
• Gate 2 — Questions → LLM with all knowledge chunks
• Gate 3 — Out of scope → Polite redirect

Anti-Hallucination

• Grounded prompt: LLM answers only from provided data
• Verification: entity consistency + overlap check
• Refusal detection: catches "I don't know" responses
• Ethical guard: blocks harmful content

---

LLM Providers

Provider	Model	Config
Simulated (built-in, offline)	simulated	Default — no setup
OpenAI	gpt-4o-mini	Set OPENAI_API_KEY
DeepSeek	deepseek-v4-flash	Set DEEPSEEK_API_KEY
Xiaomi MiMo	mimo-v2-flash	Set MIMO_API_KEY

"llm": {"provider": "mimo", "model": "mimo-v2-flash", "temperature": 0.3}

---

CLI Reference

Command	Description
kstudiochat init <path>	Create a new pack
kstudiochat init <path> --template <name>	Create from template
kstudiochat templates	List templates
kstudiochat chat --data <path>	Interactive chat
kstudiochat serve --data <path>	REST API server
kstudiochat validate --data <path>	Validate pack
kstudiochat info --data <path>	Show config

---

Multi-Tenant API (for clients)

K-CHAT can run as a multi-tenant API — clients upload their data and get a chatbot endpoint.

Endpoints

Method	Path	Description
POST	/api/packs	Upload zip → get API key
POST	/api/chat	Chat with bot (Authorization: Bearer kc_xxx)
DELETE	/api/packs/:id	Delete pack
GET	/api/widget.js?key=xxx	Widget embed snippet

Client embed (shared hosting)

<div id="kchat-widget"></div>
<script src="https://your-api.url/api/widget.js?key=kc_xxx"></script>

Direct API usage

curl -X POST https://your-api.url/api/chat \
  -H "Authorization: Bearer kc_xxx" \
  -H "Content-Type: application/json" \
  -d '{"message": "What are your hours?"}'

---

Extensions

Config-driven, enabled via config.json:

"extensions": {
  "conversational_memory": {"enabled": true},
  "predictive": {"enabled": true},
  "multimodal": {"enabled": true}
}

Extension	What it does
conversational_memory	Tracks user preferences, mood
multimodal	Enhances emotion detection
predictive	Flags risky interactions
live	Injects real-time data
adaptive	Tracks strategy effectiveness
rlhf	Collects feedback
voice	STT → chat → TTS pipeline

---

Why K-CHAT?

Instead of...	K-CHAT gives you
LangChain	Folder-based data contract, zero code
Building from scratch	Anti-hallucination out of the box
SaaS chatbots ($30+/mo)	One-time deploy, free to run
ML pipelines	Deterministic, auditable verification

---

License

MIT